code-review

A context-aware code review gathers organizational context before analyzing diffs, ensuring findings reflect team conventions. It hydrates context per area touched in a PR and categorizes findings to ground results in established patterns.

To review a multi-area PR, run a separate context research pass for each touched area before generating the review. This hydrates specific context for auth, billing, or UI, aligning findings with the relevant team patterns.

The best way to structure code review findings is using a predefined taxonomy and a report template. This categorizes findings with citations from research, producing a structured report that facilitates fast decision making.

Analyzing diffs without organizational context results in generic opinions rather than team-specific findings. This skill requires a context research pass per touched area to ground the review in actual team conventions and patterns.

For large diffs across different domains, the review process hydrates context per touched area separately. It then categorizes findings using a defined taxonomy and presents a structured report with citations to ensure comprehensive coverage.

To start a context-aware PR review, you need to provide the diff URL or PR reference. You then run a separate context research pass for each area touched before generating the final structured review report.